11 / 15 · Launch checklist

Launch checklist

Use this checklist on your customized production build, not only the sample template. Automated checks catch useful errors, while content, destinations, appearance and real integrations need browser review.

Automated checks

From the project folder with dependencies installed:

npm run format
npm run check

Review formatting changes, then wait for language/content validation, documentation, formatting, lint, TypeScript, tests and build to pass. Fix the first failed step before continuing. Run npm run start after stopping development to inspect the optimized export at http://127.0.0.1:3100.

When editing guide Markdown, also run npm run docs to regenerate the offline HTML before npm run check. Do not edit generated guide HTML independently.

Content and conversion flow

  • Replace sample branding, claims, media, testimonials, prices and names with accurate authorized content.
  • Check header, both hero actions, each pricing plan in both billing periods, final CTA, social links and the demo-access confirmation link.
  • Verify anchor links target sections that still exist. Verify external destinations and any root-relative routes exist.
  • Ensure CTA labels describe their destinations. The supplied header's kind: 'demo' action always opens demo access. With an enabled newsletter in dialog placement, exact #newsletter CTA/navigation links open newsletter signup; use such a normal link for a header intended to open signup. Other external signup/checkout links navigate directly. Review the independent dialogAction continuation.
  • Confirm annual totals, savings labels, currency, per-seat wording and FAQ amounts match.
  • Leave newsletter collection explicitly in preview, disable it, or configure a real provider/receiver. Test the full enabled journey, consent choice, confirmation, success/failure response and privacy disclosure; do not interpret a rendered field as successful email delivery.

Both language pages

  • Configure exactly two distinct language codes and the intended default; open each direct URL and refresh.
  • Review every translated section, media caption/alt text, testimonial, newsletter message, demo-access form/confirmation, cookie message and pricing line. No required placeholder should remain visibly unresolved.
  • Switch both ways with keyboard and pointer. Confirm the selected language and current section fragment, remembering that sample interaction state reloads.
  • Test root with no saved language, valid allowed memory, withdrawal and an unconfigured saved code. Explicit language URLs must always win.
  • Inspect title/description, HTML language/direction, canonical/alternate URLs and sitemap for both pages.
  • Check font coverage and translated copy at narrow widths. For a right-to-left language, review direction, mixed-script numbers, media captions and actual provider Form fields.

See Languages and translations for a complete replacement-pair example. Structural checks cannot certify translation accuracy.

Responsive layout

Review widths near 320 px and 390 px, a tablet, and a wide desktop. Also test browser zoom to 200% and any longer or translated copy.

  • No unintended horizontal page scrolling, clipped text or overlapping controls.
  • Navigation switches to the menu at 1050 px and below. Check both sides of 800 px and 1050 px, including a width between them: the menu button and its configured Get started action must remain available. Open and close the menu, then resize above 1050 px and return; it must stay closed until requested again. The main page layout retains its separate 800 px breakpoint.
  • Cards, plan prices, headings, media, gallery captions, testimonials and email fields remain readable.
  • Controls have comfortable touch targets and spacing.
  • Added images have stable dimensions and do not cause layout jumps.

Appearance preferences

  1. In a fresh test profile with no decision, confirm the banner appears and System follows the browser's colour scheme.
  2. Choose Necessary only. Select Light or Dark and confirm it works for the current page. Reload: System should return and the remembered decline should suppress the banner.
  3. Use the footer's Cookie preferences button, allow preferences, choose Light, reload and confirm Light remains selected. Repeat for Dark.
  4. Choose System with valid permission and change the system/browser colour scheme; the page should respond.
  5. Try the quick toggle, then restore System from the menu.
  6. With normal storage access, withdraw from the footer and confirm forma.theme and forma-language are removed. Reload should return to System; an explicit language URL keeps determining the page language.
  7. Test an outdated or expired decision, and blocked browser cookie/local-storage access. The page must remain usable. The detail dialog should explain if the decision only applies to this visit; a reload may ask again or restore an earlier saved choice when the new one could not be recorded.
  8. Check both palettes for readable text, buttons, focus indicators and illustration labels. Confirm an allowed saved theme is restored before paint.

The supplied decision cookie is forma-cookie-preferences with version 2. Optional theme uses forma.theme; language uses forma-language, both permission-gated. Withdrawal attempts to remove both optional values. Forma does not store newsletter addresses or consent records in browser preferences. Choosing a theme should not reset unrelated controls or shift layout; navigating to another language loads fresh interface state.

See Cookies and preferences for the full inventory, privacy-link configuration and migration behaviour. Inspect any services you add separately: this feature does not automatically block third-party scripts or establish legal compliance.

Keyboard and accessibility

  • Use Tab and Shift+Tab to reach controls; every control should have a visible focus indication.
  • Use the mobile menu, theme options, pricing switch and FAQ without a pointer.
  • Use the language switcher without a pointer; confirm both native language names and accessible destination labels.
  • Use screenshot thumbnails and previous/next buttons without a pointer. Check the selected image, position/caption and alternative text. One configured image should have no gallery controls.
  • In a test copy, rebuild with siteSettings.preview.video.frameStyle set to 'window', then 'minimal'. Window mode should show its decorative browser header and bottom descriptive caption bar; minimal mode should remove the header's three dots, product name/top label and the bottom descriptive text bar without leaving an empty caption bar. Both must keep the rounded frame, outline, ring and shadow, including the responsive phone treatment. The descriptive caption must remain available to screen readers, and the separate sample-media disclaimer must remain visible. Both must fit phone/tablet/desktop widths without overflow and retain the same keyboard, playback, loaded track captions and native/error fallback behavior in both themes. Verify a playback-error message is still visible on failure. Omitting frameStyle must retain the window presentation. Return to image mode to confirm its gallery is unchanged, then restore your intended production settings and rebuild.
  • For video mode, click or tap the video surface to play and pause, then repeat using Tab and Enter/Space. Focus should remain on the surface during playback. Check the brief Play/Pause feedback and its disappearance. The paused centered Play affordance returns; at 600 px and below, it hides while the bar is visible and returns when the paused bar hides. Brief phone feedback appears above the bar. Test the bar's Play/Pause control separately, seeking with arrow keys, elapsed/total time and replay after the video ends. Review both themes and language pages; translate every control label.
  • In the page, leave the pointer idle during playback for more than three seconds and confirm the floating bar hides. Mouse/pen entry or movement should reveal it. While paused or ended in the page, hovering should keep the bar visible; moving the pointer away should hide it even when stopped. A touch tap should give about three seconds of in-page visibility whether playing or paused; lifting a finger alone must not hide it immediately. In supported fullscreen on desktop and mobile, enter by mouse/touch and wait without moving: the bar must hide after about two seconds whether playing, paused or ended. Movement, a tap or release of a held control should restart that timer. Hold the seek or volume control still for longer than two seconds: the bar must remain visible until release/cancellation; also test releasing outside the player and switching away from the window. A pointer click or focus restored by entering fullscreen must not pin the bar indefinitely. Exit fullscreen and recheck the original in-page behavior. Repeat with Tab/Shift+Tab entering from outside the player and keyboard-focused controls: the bar must stay available in either view.
  • At 600 px and below, verify the compact two-row layout: seek/time share the 24 px-high top row, with available 44 px control buttons grouped at the start of the second row with 8 px gaps. An absent captions or fullscreen button must not leave a reserved empty slot. Test 320 px with captions loaded and fullscreen supported; the bar must stay within the video stage without extra wrapping, clipping or horizontal overflow. Repeat after changing video aspect ratio. The seek row is intentionally smaller than the buttons.
  • Test mute/unmute and the volume slider with your actual audio. At 600 px and below the slider is hidden; device controls may determine sound level. The supplied silent clip cannot demonstrate audible volume changes, so use an authorized audio-bearing test clip when replacing it.
  • With a valid caption track, check its content and the captions toggle after loading; without a track there should be no captions button. Review captions with the floating bar visible and hidden at phone widths, fullscreen and 200% zoom. Supported automatic bottom cues should move above the visible bar and return when it hides; explicit positions in your WebVTT file remain authored and need separate review. Browser caption preferences/support may change placement. Test fullscreen where the standard browser API is supported, exit it, and confirm playback remains usable if fullscreen is unavailable or denied.
  • Confirm no unwanted autoplay, preload="none", stable poster dimensions and no layout shift during loading. Check reduced motion disables the small playback halo and keeps the brief centered feedback static.
  • On the production host, verify correct video/caption media types and video HTTP byte-range support. After Play, seek forward/backward and confirm time/playback reach the requested position; successful initial playback alone does not verify seeking. Compare your clip with the supplied local production helper if the live host behaves differently. See Build and deploy.
  • In a separate test copy, use an unavailable video path to check the translated playback-error message and restored native controls. The custom retry bar should move below the video, leaving the native controls unobstructed. If the enhanced surface had focus, it should move to the bar's retry control. Restore the correct path before publishing. Check native controls before the enhancement and with JavaScript disabled too; Forma's themed bar requires JavaScript.
  • Test testimonial previous/next and Pause/Play. Hover/focus and a hidden tab should pause autoplay; reduced motion should disable it. With zero entries the section hides; with one, controls hide. Review the longest quote in both languages for stable height.
  • With demoAccess.mode: 'preview', test Get started using sample data. Empty name/email, a spaces-only name and an invalid email must not continue. Valid button/Enter submission clears fields and focuses the labelled demo confirmation; its explanation must say no account, email send or saved/transmitted details. Check the configured public continuation. Close from each step with the close control, Escape and backdrop; reopening must reset the form, with appropriate return focus. Check phone scrolling, keyboard, both themes/languages and 200% zoom. Confirm no provider request or input storage. A public continuation is not authentication.
  • On invalid demo-access, preview, native Kit and enhanced endpoint submission, verify translated inline errors, invalid states/descriptions and focus on the first invalid field. Correcting a field should clear only its own error. For newsletter, test empty/invalid email and unchecked required consent separately and together. Check checkbox contrast, focus and forced-colors. Endpoint without JavaScript retains native constraints and POST. Preview/native newsletter controls remain disabled without JavaScript; native signup has a hosted fallback. Embedded Kit's validation needs separate provider review.
  • With enabled newsletter placement dialog, use exact #newsletter actions to open signup and kind: 'demo' to open independent demo access. Test Tab/Shift+Tab, close/Escape, return focus, scrolling and 200% zoom. Closing/reopening newsletter should preserve its instance; switching the shared dialog to demo access should replace it. Demo access must reset on close regardless of newsletter placement. Ordinary external signup/checkout links must navigate directly.
  • At phone widths, open the mobile menu and use its Get started CTA. It must use the same configured destination and translated label as desktop Get started, while the hero action remains independent. After opening the popup, the menu should close; dismiss the popup and confirm focus reaches the menu button rather than a hidden menu item. On desktop, focus should return to the visible opening control. In newsletter popup placement, dismissing its inner preview confirmation must leave the outer signup dialog open and return focus to newsletter submit.
  • For the cookie-preference dialog, test Allow preferences, Necessary only/Withdraw preferences, close, Escape and return focus. Review the banner and dialog at narrow widths and 200% zoom.
  • Confirm headings are logical, icon-only controls have meaningful names and links have useful labels.
  • Review informative image alternative text and decorative graphics' quiet treatment.
  • Check reduced-motion behaviour and that information does not depend only on colour or hover.

An automated accessibility scan is helpful but does not establish complete conformance. Repeat manual review after custom layouts or integrations.

Performance review

Measure the production page using browser tools or Lighthouse with a consistent device and network profile. Development mode performs extra work and is unsuitable for final speed assessment.

Track loading speed, layout stability and interaction responsiveness. A local benchmark or Lighthouse score alone cannot guarantee field Core Web Vitals. Use the current Core Web Vitals overview for metric definitions and assessment criteria.

Target LCP ≤ 2.5 seconds, INP ≤ 200 milliseconds and CLS ≤ 0.1 at the 75th percentile of real visits, assessed separately for mobile and desktop. A field pass needs evidence for all three; a new site with insufficient field data must report that limit. Lab TBT is useful for diagnosing loading work but does not establish field INP. Keep slower runs and warnings when comparing production changes.

Preserve the template's lightweight approach: limited client-side controls, local font, optimized local media and static hosting. Compress added images/video, reserve asset dimensions and avoid loading large libraries for small effects. The supplied screenshots/poster use WebP; their SVG originals remain editable. Review a replacement's compressed size and readability rather than relying on its file extension alone.

In the production network panel, reload with an empty browser cache and inspect the configured initial image/poster. Landing pages preload the video poster, or the first product image when it has no darkSrc; the selected screenshot loads eagerly with high priority, while later feature/workflow images and thumbnails are lazy-loaded. Confirm your replaced paths resolve without errors and scroll through later images to check they appear. Do not add high-priority preloads for every image.

Check the local font on that fresh load too: its generated WOFF2 URL should resolve, start early and transfer once rather than duplicate a differently configured preload. The post-build helper supplies a missing preload for the unambiguous single-font export without a buyer setting. If you replace the font or add more fonts, inspect the generated page head and actual requests rather than assuming this single-font helper chooses them for you. Review text after font loading and retain the appropriate notices.

If a replacement font has different character coverage, update or remove the old unicode-range in document-frame.tsx and confirm the new glyphs actually use it. Test fallback text as well as Latin content; a successful font request alone does not establish correct character selection or a stable layout.

Later sections use CSS content-visibility: auto with estimated heights, while their content stays in the document. After adding cards, gallery rows or long copy, review the --section-estimate rules in Appearance. Check these paths on both language pages and at phone/desktop widths:

  • Scroll down and back up; sections should appear without clipping or unexpected jumps.
  • Open a section's direct fragment URL in a fresh load, then use header/footer anchors and browser Back/Forward. The intended heading should be reachable and readable below the sticky header.
  • Use Tab and Shift+Tab through later controls and browser Find for text in an initially offscreen section. Confirm the matching content becomes visible; repeat accessibility review with the assistive technology you support.
  • Inspect print preview for all sections. Print styling removes deferred rendering and estimated section sizes.

Supporting browsers may defer offscreen rendering; unsupported browsers render normally. This CSS does not remove the semantic content, replace manual accessibility testing or guarantee a particular performance score.

The video uses preload="none" and does not autoplay. Measure the initial page separately from user-requested playback and inspect transferred bytes after Play. Re-measure after adding third-party services. Embedded Kit loads on section mount or popup opening by default; compare that state with optional interaction-gated loading. For native Kit, compare the initial page with a valid opted-in submission that initializes the public handler. Provider requests and interaction costs are part of the configured real experience.

Test both language pages on a throttled mobile profile as well as desktop, then repeat on the intended HTTPS host. Hosting compression, cache headers, network distance and third-party services can change results. Repeat comparable cold-load runs rather than treating one favorable run as a guarantee for every visitor. If a result misses your launch target, inspect its loading and main-thread work before publishing.

Record the URL, date, browser/tool version and testing conditions so future changes can be compared fairly.

  • Before real collection, the buyer/site operator must configure and test suitable provider-side spam/abuse controls or protect their own receiver. Forma includes no server-side anti-spam service; browser validation, consent controls and double opt-in alone cannot prevent automated submissions. Keep required Kit branding visible and obtain provider clarification before production use of the advanced custom adapter.
  • In preview mode, test invalid email and unchecked consent, then valid sample submission by button and Enter. Confirmation must explicitly say it is a demo with no email, subscription or saved/transmitted details. Fields reset; close/Got it/Escape returns focus to submit and another attempt works. Verify no provider request or preview input storage. Without JavaScript, preview controls remain disabled. Review both themes/languages and phone/tablet widths.
  • Test section and popup placements. In the newsletter popup, the nested preview confirmation must leave the signup popup open when dismissed and return focus to submit. In disabled mode, hide signup and remove or redirect its links.
  • For embedded kit mode, test automatic loading at section mount or first signup-popup opening. If kitLoad: 'interaction' is configured, verify no embed request before Load signup form. After a real Form renders, that optional button must disappear; focus should enter it only if the loading control still owned focus. Check actual provider fields, unchecked consent, privacy links, confirmation, correct tag/status and unsubscribe using an authorized owned address.
  • Publish each embedded Form with Advanced return-visitor behavior Continue to show the same Form. Revisit after signup in both the same and a fresh context. Test unpublished/cached changes, blocked scripts, Retry and hosted fallback. The builder can update before the public embed; wait for provider propagation rather than masking a cache delay in Forma.
  • For kit-native, confirm Forma's email/required-choice fields appear immediately without a Load/ready button. Invalid or unchecked submission must initialize no handler and send no signup. A valid selected choice should submit once, retain provider verification steps and show an honest pending/received/error/unknown response. Received must not claim confirmed subscription. Verify the actual Form, permission tag, confirmation and unsubscribe in Kit. Test hosted fallback, script/network failure, a verification step when encountered and unresolved requests; never report an unencountered real challenge as passed.
  • For demoAccess.mode: 'kit', first submit valid name/email with optional marketing unchecked: continuation must appear with no Kit signup request. Reopen and test a selected choice: the separate Form must receive the intended name/email/tag, while the continuation remains usable independently of signup outcome. Check close/reopen, duplicate activation, failure/unknown response and confirmation/unsubscribe. These are mailing-list tests, not authentication.
  • With the standard Kit embed and native demo signup enabled together, start a request that is awaiting verification, then try the other form. Repeat in the opposite order. Only the first request may reach Kit; the blocked attempt must show the translated busy message and configured hosted fallback, while demo continuation remains available. Include repeated submission in the same form.
  • In both embedded and native modes, open verification while the Form is inside a Forma popup. Its actual provider controls/iframe must stay interactive above the popup, with useful focus and the original provider callbacks. Check the provider Close control and Escape; neither may fabricate completion or bypass verification. Include closing the original popup where that journey permits it. Offline fixture results do not establish a real CAPTCHA pass.
  • Keep the first result unresolved through the unknown-state timeout, close/reopen the popup and switch its purpose. None should release the shared request lock or create a second verification request. Settle the original request with a late provider completion/error after its original view is gone; confirm it releases the correct request and cannot make a blocked second request appear received. Retry only after the first resolves. Use isolated synthetic verification for repeatable overlap checks; do not call that real confirmation or delivery acceptance. Page-local coordination does not protect across reloads or tabs.
  • Both stable language slots need the appropriate public Form settings. Translate native Forma notices/status/errors in the dictionaries; translate embedded fields and provider confirmation in Kit. Reusing one English Form deliberately does not translate its provider-owned content.
  • In a newsletter popup, unfinished entries stay through close/reopen unless the shared dialog switches to demo access. Refreshing or changing language starts fresh interface state. Closing or switching cannot cancel a sent request, unload Kit, revoke provider storage or withdraw email permission.
  • In endpoint mode, verify exact field mappings/values, required constraints, receiver validation, response/redirect and failure handling. The receiver must timestamp, retain/protect records and handle withdrawals; it is not included.
  • Review actual services/storage and keep email permission separate from theme/language memory. For embedded Kit, an unchecked provider checkbox may still allow signup; target marketing only to active/confirmed subscribers with the intended permission. A tag can remain after unsubscribe and is not sufficient by itself. Native Kit submits the selected tag, but Forma's notice-version/source values are not automatically forwarded as a consent ledger.

Follow Email signup and Kit for setup and complete test instructions. Real tests require the owner's authorization for the recipient and destination. Keep personal data, confirmation tokens and private API/webhook credentials out of public configuration and test artifacts. Record untested delivery, provider challenge, translation or accessibility journeys as remaining checks.

Security headers

  • Build the intended configuration, run npm run security:check, then preview with npm run start. The security check compares current settings and exported HTML with the generated manifest and _headers. Preview checks the requested document's digest; it does not recheck edited source settings or the full header recipe at startup. Resolve mismatches with a fresh build; do not bypass the check or edit hashes.
  • Inspect Content-Security-Policy on the root, both language pages and the missing-page response. Confirm script hashes are present without script unsafe-inline or unsafe-eval, and check the browser console while exercising the page.
  • Test theme/language startup and switching, cookie allow/decline/withdrawal, menus, forms, video playback/seeking/captions and keyboard interaction under that policy. Repeat after adding a service or modifying inline scripts.
  • With both collection modes in preview, verify there are no external Kit permissions or requests. For an active integration, check its generated script/connection/frame/form permissions and its actual verification/fallback journey. Review custom provider fonts/analytics separately, along with the conditional reCAPTCHA rules. An allowed request alone is not a confirmed subscription.
  • Verify remote configured images and media have their exact origins in the policy, including dark variants, avatars, posters and caption files. There should be no blanket HTTPS image/media permission. Check any custom-coded assets against the extra-origin configuration and confirm workers remain blocked unless you deliberately design a separate reviewed change.
  • Confirm the actual static host applies the generated production headers. A host that ignores _headers needs its own configuration; a successful local preview is insufficient. Publish HTML/assets/headers from the same build and review extra CSP headers or CDN-injected script changes.
  • On a valid HTTPS deployment, verify Strict-Transport-Security has the intended lifetime. The default is 30 days without subdomain coverage or preload. Ordinary HTTP localhost should have no HSTS header. Test HTTPS and redirects separately before enabling persistent HSTS on a real host.

See HSTS and Content Security Policy for configuration and hosting boundaries. These checks do not establish full security certification or replace provider/backend abuse controls. Repeat comparable performance checks after configuration and hosting changes.

Production and metadata

  • Set NEXT_PUBLIC_SITE_URL before building and confirm your canonical/social URLs are correct in page source.
  • Confirm title, description, language, favicon and 1200×630 social image match the site.
  • Confirm direction, alternate links and sitemap match the configured language pair.
  • Inspect production console and network panels for unexpected errors or failed assets.
  • Confirm your privacy-policy link exists, preference labels/inventory match actual storage, and any added services have their own tested controls and disclosures.
  • Check refresh and a missing-page URL on the intended host.
  • Retain runtime/font licence notices with the compiled files.
  • Keep credentials, source archives and private environment files out of the public document root.

For the final upload sequence, use Build and deploy. For common failures, use Troubleshooting.